You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Predicting mid-to-mid price change over ~100 ticks is table stakes for a big quant fund (and easy even without ML). If you can monetise the prediction after accounting for latency, commissions, bid/offer spread and slippage, *then* I’ll be interestedhttps://www.bloomberg.com/news/articles/2021-06-08/man-group-oxford-quants-say-their-ai-can-predict-stock-moves …
They do say they intend to use it for execution, not prop trading. But still — the question is not whether your neural network can predict stock prices. It’s whether it can do it better than the existing linear models that your execution team already uses.
Saying you use AI gets you lots of publicity though.
What if I trade only in those stocks where the mid-to-mid predicted price change is high enough for the costs to be amply compensated and then some?
Sure, if you can reliably identify those stocks in advance.
I think the article makes most sense when read as marketing for the ML-optimised hardware company involved in this project.
Yeah it’s clearly marketing — for MAN, for the research institute and for Graphcore. Doesn’t mean we shouldn’t call out the grandiose claims tho!
I sensed that this was, albeit cool, a bit of a hope project. Manager would ask, slippage? size movement affecting price? actionable beyond a cool backtest? To which the answer is... 'gotta check something, brb.'
In my experience it's much easier to do stock classification than regression (which is what's the NN achieves). Concept drift is also very frequent in regression since stock prices tend to regress to their mean value
Binary rv's have a thin tailed distribution, so they are inherently easier to predict wrt returns
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
